Evaluation of Muscle Mass and Quality With an AI‐Based Muscle Ultrasound Imaging System in Patients at Risk of Malnutrition

TL;DR

An AI-based ultrasound system was tested to assess muscle mass and quality in malnourished patients, showing that it can detect sarcopenia-related changes effectively.

Contribution

The study introduces an AI-assisted ultrasound system for evaluating sarcopenia in malnourished patients, demonstrating its effectiveness in measuring muscle mass and quality.

Findings

Patients with sarcopenia had significantly lower muscle thickness and area compared to those without sarcopenia.

AI-assisted ultrasound parameters like muscle thickness and low echogenicity were significantly lower in sarcopenic patients.

Higher muscle thickness was protective against sarcopenia, while higher low echogenicity was protective against low handgrip strength.

Abstract

Sarcopenia is characterized by the loss of muscle mass, quality and function. Ultrasonography provides a non‐invasive method for assessing sarcopenia. Its generalizability remains limited due to certain methodological and population‐specific challenges. This study evaluated the association between AI‐assisted muscle ultrasonography and sarcopenia in patients at risk of malnutrition.

This observational, cross‐sectional study included 647 patients at risk of malnutrition. Nutritional status was assessed via anthropometry, bioimpedanciometry, quadriceps rectus femoris (QRF) ultrasonography and handgrip strength. An AI‐based imaging system segmented the region of interest (ROI) in transverse QRF images to measure muscle thickness (RFMT), area (RFMA) and pennation angle (RFPA). The Multi‐Otsu algorithm extracted ROI biomarkers: low echogenicity (MiT) and medium echogenicity (FatiT), assumed as a surrogate of muscle and fat percentage of the ROI. Sarcopenia was diagnosed using European Working Group on Sarcopenia in Older People (EWGSOP2) criteria and malnutrition was assessed with Global Leadership Initiative on Malnutrition (GLIM) criteria.

Most of the patients of the study were female (54.4%) and the mean age was 64.83 ± 15.79 years. Malnutrition was present in 530 patients (81.9%) and sarcopenia in 167 patients (25.8%) Among patients with sarcopenia 57.2% had low muscle mass, and 44% had low handgrip strength. Patients with sarcopenia had significantly lower values of RFMT (sarcopenia: 0.89 ± 0.27 cm; no sarcopenia: 1.03 + 0.29 cm; p < 0.01) and RFMA (sarcopenia: 2.77 + 1.02 cm²; no sarcopenia: 3.25 + 1.17 cm²; p < 0.01). In terms of muscle quality by AI‐assisted ultrasonography, we observed lower values of pennation angle (sarcopenia: 4.97 ± 2.91°; no sarcopenia: 5.50 ± 2.78°; p < 0.01), low echogenicity (MiT) (sarcopenia: 45 ± 10.80%; no sarcopenia: 47.39 ± 10.91%; p = 0.02) and a higher high echogenicity percentage (NMNFiT) (sarcopenia: 14.99 ± 5.52%; no sarcopenia: 14.76 ± 5.17%; p = 0.02). Multivariate analysis showed male sex as a risk factor for sarcopenia (OR = 1.85 (IC 95%: 1.23–2.77); p < 0.01), while higher RFMT was protective (OR: 0.18 (IC 95%: 0.04–0.86); p = 0.03). For low handgrip strength, higher MiT was protective (OR: 0.07 (IC 95%: 0.13–0.43); p < 0.01) after adjusting for age and sex.

In patients at risk of malnutrition, sarcopenia and dynapenia were associated with reduced muscle mass and quality. AI‐based ultrasound parameters, particularly RFMT and MiT, were significantly lower in individuals with sarcopenia and correlated with poorer muscle function, independent of age and sex.

Figures